Is a time series model the most appropriate to answer my questions

#1
I would like to know if I should analyze my data using a time series model (ARMA,...).

Context: I want to test a detailed language grid-analysis that I have developed to analysis verbal communication between 2 individuals. The topic of the conversation is predefined and the interlocutor is the same for each subject. Thus the same "actor" will converse with all the subjects on the same topic.

Data: My dataset is organized by turn of speech. By turn of speech, I mean each time one speaks. My data is basically the occurrence of tagged groups of words per “turn of speech” throughout the dialogue (around 300 turn of speech per dialogues). I am looking at two distinct words categories (A and B), each of these 2 categories have 3 sub-categories (A1,…A3 and B1,…B3). I want to see 1) A1 and A3 occur as often as B3; 2) A2 occurs as often as B1 and B3.

Occurrence of words
Turn of speech A1… A3 B1... B3
1 0 0 1 0
2… 0 4 3 0
300 1 2 0 2

I am doing a test on a sample of 4 different dialogues before pursuing the analysis on a bigger sample (20 dialogues). To adjust for the length difference amongst the dialogues, I am using frequency scores and dividing the occurrence per total number of speech.

Problem: I oriented my research towards time series analysis but have a doubt if this is the right analysis for 2 reasons: 1) I am not really trying to predict the course of a new communication pattern but to understand the profile of the current ones. 2) I have 2 variables (A and B) changing over time and not 1 and I have the impression that time series analysis are done on a single dimension. Should I opt for another strategies?

I thank you for any help you can provide and look forward to your response.